Tree-ring width (TRW), and even more so maximum latewood density (MXD) chronologies, have become the backbone of summer temperature reconstructions and the assessment of the climatic impact of past volcanic eruptions . To put the cooling induced by the three 17th century volcanic eruptions into perspective, we reconstructed summer (i.e., June–August, or JJA) temperature anomalies for Fennoscandia and Ostrobothnia using multi-centennial Scots pine (Pinus sylvestris L.) chronologies (MXD and TRW) from four sites located in northern Fennoscandia (Fig. ) . Each chronology was standardized using the regional curve standardization (RCS) method as it optimally preserves multi-decadal temperature variations in the reconstruction . We then transferred this record into JJA temperatures through a bootstrap linear model using a principal component analysis (PCA) calibrated against JJA land surface temperatures (1920–2000) from the E-OBS 10 min × 10 min gridded dataset . For each grid point, the calibration and validation process was repeated 1000 times using a bootstrap approach. To test the robustness of the reconstruction, we employed the coefficient of determination (R2 for the calibration and r2 for the verification periods), RE (reduction of error), and CE (coefficient of efficiency) statistics (Supplement, Fig. S1).

As the main livelihood of the majority of the study area's population came from agriculture, and specifically from crop cultivation, the possible impacts of volcanic eruptions on livelihoods are studied here with grain harvest proxy data. Grain tithe tax records are commonly used to detect annual harvest fluctuations in historical Europe , and tithes have been shown to reflect relative harvest fluctuations rather well, particularly in early modern Sweden . In 17th century Ostrobothnia, each peasant holding was supposed to pay a share (ca. 10 %) of their yearly harvest as a tithe tax to the authorities . The tithes consisted of roughly equal shares of rye and barley, but in the northern region, tithes were paid almost entirely in barley. Previously published tithe series from southern Ostrobothnia were extended to cover the whole study area. The tithe time series were collected on a parish level and then transformed to indicate annual variations with respect to a pre-crisis (or pre-peak forcing) 10-year mean.

The grain tithe time series were analyzed with a superposed epoch analysis (SEA) to investigate the possible volcanic signal in the data . As the volcanic events included at least one unidentified eruption (1695), the time series were superposed on the year of peak global volcanic aerosol forcing rather than on the year of the eruption. The dating of peak forcing year is based on . In addition, spatial variations and relationships between 3-year mean harvest losses were explored at the parish level.

The main economic burden for the 17th century peasant was paying taxes to the Crown. Thus, the possible socioeconomic repercussions are assessed here with tax debt records. These records are particularly appropriate to address the societal and economic consequences on a grassroots level, as the majority of the people living in the studied area were peasants, their family members, or farmhands living with the peasant family. Thus the tax-paying ability of a holding reflects the overall subsistence status of the extended family household.

If a farmstead within the 17th century Swedish kingdom failed to pay taxes over 3 consecutive years, it was marked as öde, which means “deserted”. Hence, the holding was not always deserted in a literal sense if it was marked as such (Fig. ). Nevertheless, we will use here the literal English translation of öde to avoid possible anachronistic expressions, like insolvency. In addition, the category of deserted holdings included uninhabited farmsteads as well, because a holding was also marked as deserted if the peasant fled, migrated, or perished . Thus, these desertion markings overall serve as a proxy for subsistence decline, and in many cases for severe societal hardships. Furthermore, desertion variations can imply changes in land tenure and social status. If a freeholder farmstead was marked as deserted, the peasant lost its hereditary right for the property, and it was released for the Crown to claim. In the best case, the peasant could continue cultivating the farm as a tenant of the Crown – i.e., as a Crown peasant – without the hereditary and usufruct rights for the holding. In the worst case, the family was evicted from the farm. The situation was even more perilous when a Crown peasant had tax-paying difficulties, as the authorities did not recognize their usufruct rights. Consequently, these situations commonly ended with eviction . In a rural society like that of 17th century Ostrobothnia, such evictions inevitably affected the economic possibilities and social status of a person.

The desertion material was retrieved from various accounts related to taxation, which are bound as verification, account, and land books (Fig. a). These records are available in so-called bailiff (up until 1634) and provincial (from 1635 onwards) accounts, stored at the National Archives of Finland. The number of total and deserted holdings are given in mantal units in the documents (Fig. c). Mantal was a Swedish farm taxation unit, which was assigned according to the quantity and quality of the arable land of a peasant holding . The desertion data were collected for the 7 years preceding and the 7 years following peak volcanic aerosol forcing. When available, the share of freeholder and Crown peasant holdings (both the total number of both holding types and the share of deserted holdings) was also gathered. The collected time series indicate the average desertion ratio, that is the share of deserted holdings in relation to all holdings (based on the number of holdings existing in the year of peak volcanic aerosol forcing).

The desertion ratio data were examined with the SEA procedure similar to the grain tithe data. However, whereas the grain tithes were recorded on a parish level throughout the 17th century, the number of deserted holdings were documented on spatially wider administrative units in the years 1594–1599 and in 1602 in the sources. These wider areas are not comparable with each other regarding the number of farmsteads (varying from 122 to 805 holdings). Thus, to avoid a possible bias towards more sparsely populated areas in the analysis, the desertion ratio SEA was performed for five comparable sub-regions (Fig. ). Additionally, the 3-year mean desertion ratios were explored at the parish level.

The 17th century saw a series of major volcanic eruptions – according to the volcanic explosivity index (VEI) and using a threshold of VEI≥5, between two and seven eruptions comparable to 1601, 1641, and 1695 would have occurred in the 17th century (see, e.g., ). By contrast, even listed as many as nine VEI≥5 eruptions that would have contributed to the climatic regime of the 17th century global crisis besides the 1600, 1640–41, and 1695 events. In the most recent database, list 12 eruptions for the 17th century, but they also state that nine of these had a limited climatic impact as their estimated volcanic stratospheric sulfur injection (VSSI) was <5 Tg[S]. Only the three eruptions in 1600, 1640/41, and 1695 released more than 15 Tg[S], and thus likely triggered substantial cooling over the study area. In addition to the absence of a linear relation between the amount of sulfur injected to the stratosphere and the resulting temperature cooling, one should also consider the prevailing background climatic conditions as well as eruption location and season , as they will critically determine the climatic effects of a given eruption. Many past studies solely used the VEI to link climatic or societal deterioration to volcanic forcing, even if it is well established that this index does not directly indicate the climate impact potential of a volcanic eruption . We can thus only insist that any direct attribution of societal repercussions to volcanically forced cooling using evidence of VEI estimates alone should be avoided.

Whereas the cooling induced by some of the other 17th century eruptions cannot therefore be attributed easily to volcanism alone, we evidence that the VSSIs and their impact on global radiative forcing likely contributed considerably to the cold summers observed over Fennoscandia in 1601, 1641, and 1695. The strong JJA cooling shown in this study (Fig. ), especially in 1601 and 1641, is in agreement with observations and model simulations , indicating pronounced post-volcanic summer cooling over most of Europe. Considering the unidentified 1695 volcanic eruption, the very cold JJA temperatures reconstructed in 1695 can be associated with volcanic forcing if the eruption has occurred in summer or autumn 1694. Should the unidentified eruption instead only have taken place in winter 1694–1695 or thereafter, one could not exclude that the cold conditions in summer 1695 resulted simply from internal climate variability, which produced several cold extremes in the absence of volcanic aerosol forcing during the Maunder Minimum.

The grain harvest and socioeconomic responses lasted considerably longer than the volcanic-induced cold pulses (Fig. ). The prolonged impact on grain harvest is likely partly explained by the poor quality and quantity of seed grain and partly because of the continuing unfavorable (non-volcanic) weather conditions, or the combination of both . For example, the seed grain aid imports from the Baltics arrived too late for sowing in 1697 because ice clogged the Ostrobothnian harbors long into spring or because the ships were wrecked in storms . Also, whereas the onset of successive years of poor harvests in 1601 and 1641 coincides with the estimated peaks in global radiative forcing and SAOD , this tight temporal correspondence between the estimated SAOD and the cold summer 1695 reconstructed over Ostrobothnia is not evident (Fig. ). Thus, we cannot rule out the possibility that factors other than volcanic aerosol forcing may have contributed to the degree of harvest losses in that particular year. Indeed, dry summer and early autumn frosts damaged the harvest already in 1693 . Furthermore, in addition to summer season temperature variability, pre-industrial crop cultivation in Finland was sensitive to winter severity and the related onset of the growing season . In years following long, cold, and snow-rich winters, the onset of the new growing season was badly delayed, and the crops may not have had time to ripen before the occurrence of the first autumn frosts . Notably, winter temperatures in Finland are influenced partly by the intensity of westerly airflow, and the latter is linked to the modes of the North Atlantic Oscillation (NAO) . The years of negative winter NAO phases are strongly correlated with cooler winter temperatures and thicker ice and snow cover, with likely impacts including a later onset of springs and early summers due to ice–albedo feedbacks . A documentary-based NAO reconstruction indicates that the winter (i.e., December–February) 1694–95 NAO was among the most negative (-2.56; first percentile) over the period 1500–2000. Furthermore, ice break-up dates from the Torne River (65.84∘ N, 24.15∘ E) provide further evidence for an extremely cold winter and spring, as the ice broke up as late as 5 June 1695 , i.e., at the second latest ice break-up date recorded since observations started in 1693. In the year of the latest ice break-up, 1867, extreme harvest failures and hunger were also witnessed in Ostrobothnia . Thus, the substantial harvest losses in 1695 cannot likely be explained exclusively by the cold summer, but also by the extremely cold winter 1694–95 and the resulting delayed onset of the growing season, which may have been unrelated to the volcanic aerosol forcing.

The socioeconomic consequences varied considerably over time, space, and the peasant community (Fig. , Table ). The moderate socioeconomic consequences following the 1695 unidentified eruption are rather unexpected in particular, as it is well established that the demographic effects of this crisis were devastating. One of the most calamitous famines in European history raged in Finland in 1696–1697, when over one-quarter of its population perished . Furthermore, although the majority of the population gained their livelihood from agriculture, the severity of harvest losses did not always dictate the socioeconomic outcomes (Fig. ). Thus, some other factors must have influenced whether the volcanic eruptions had strong or weak societal effects over different periods and locations, or among different societal groups. In order to reveal these influencing factors, we need to investigate the socio-environmental context with high spatial and temporal precision.

Following the 1601 peak in volcanic forcing, the parishes with greatest harvest losses had the highest share of deserted holdings (Fig. a, d). This indicates that the crop failures had a direct influence on the socioeconomic conditions. The reason for this is likely connected to the overall stability of the society. The Swedish realm was in great distress at the turn of the century. The kingdom was at war with Russia, and Russian military raids caused destruction over northern Ostrobothnia . Within the realm, King Sigismund and Duke Charles fought over the throne. These struggles materialized among the people in Ostrobothnia as well, as taxes increased and the peasants were obliged to provide maintenance and provisions for the troops passing through . Furthermore, besides the increased fiscal burden and plundering soldiers, several crop failures emptied the grain stores in the 1590s. This political and military distress escalated with a peasant uprising in 1596–1597 in Ostrobothnia and a civil war in 1598–1599 in Sweden proper . The turmoil likely fundamentally decreased the capacity to cope with the 1601 crop failure: institutions for providing aid were paralyzed, trade was disturbed, and grain could not be shipped from less affected regions . The society was overall in a vulnerable state, and the troublesome times decreased the resilience of individuals. These factors likely explain why the harvest losses directly influenced socioeconomic conditions during the first volcanic event.

The direct spatial correspondence between harvest losses and desertion ratios weakens over the 1641 crisis (Fig. c, d). Furthermore, the share of deserted holdings starts to rise already in the first year of bad harvest (Fig. b). These matters imply that some other factors may have influenced the relationship between climate variability and socioeconomic consequences. The Swedish realm entered the Thirty Years War in 1630. The material and human resources for the war were raised mostly from the peasant society by increasing taxes and by conscripting peasants as soldiers, especially from the late 1630s onward . Thus, the main reason for the recorded tax debts in the 1640s may not have been linked to harvest failures, but to the raised overall tax burden. Also, the peasants likely had less grain for storing because of the raised taxes and other financial burdens of wartime, which may have negatively influenced individual coping capacity. Furthermore, the wartime may have also influenced agricultural productivity, which relied on man- and horsepower . During the years 1638–1649, altogether 4155 men were shipped from the study area to fight the war on the continent . Thus, on average, almost each household was missing one adult man contributing to the labor supply, as there were approximately 4500 peasant holdings in the area at the time. This lack of manpower may have also hindered the correct timing of farm duties, which was crucial to secure a sufficient harvest .

The harvest losses in 1695–1697 were dramatic across the study area, although the southern parishes seem to have coped slightly better than the northern ones (Fig. c). These parishes might have benefited from the grain magazines maintained by the regional authorities, where peasants could loan seed grain . However, the mitigating impact of the southern granaries was likely rather marginal as even the less affected regions lost more than half of their harvest on average.

Despite the devastating harvest losses of the 1690s, the socioeconomic impacts were rather moderate when compared to the earlier crises (Fig. d). By the end of the century, Sweden had risen as one of Europe's great powers and was preparing for a war against Russia. The economy of the realm was largely financed by taxes and tributes paid by the peasantry . Thus, the Crown could not afford to evict masses of peasants due to unpaid taxes, as in this sparsely populated kingdom there would not have been enough new people to take over the deserted farmsteads . It was more reasonable for the realm to lose tax incomes for some individual years, instead of farmsteads being permanently deserted and not producing any tax payments in the future. Consequently, the Crown responded to the 1690s crisis with unusually extensive agro-economic relief measures in the form of tax reductions and deferments . Furthermore, the authorities organized additional grain shipping and issued exemptions for Baltic Sea trade so as to increase the amount of imported grain . Yet, it is likely the volume of shipped grain was not sufficient, as the main export regions – like the Baltic provinces – also suffered from severe harvest losses . Besides, the grain imports helped people living close to seaside towns and along the main transportation networks , but the measures did not reach regions with poor road networks, like the central and easternmost regions of the study area (Fig. d). More importantly, not everyone was entitled to the shipped grain, as it was not given out for free, but either sold or given as a loan . Unlike in many western European regions, where formal poor-relief institutions and charity organizations helped the poor to mitigate and cope with crop failures , such formal institutions barely existed in early modern Sweden – and especially in its rural peripheries like Ostrobothnia. Thus, the rural poor were largely excluded from any institutional relief actions during the 17th century crises.

Although the socioeconomic consequences of the 1690s crisis were not substantial, they had severe demographic consequences. The extensive harvest failures triggered one of the worst famines in European history. In 1697–1698, population numbers dropped by 18 % in the southern parts and by 34 % in the central and northern parts of the study area . The high mortality cannot, however, be explained solely by starvation per se, but by the spread of fatal hunger-related epidemics . As there were no effective official poor-relief institutions, food aid in 17th century Ostrobothnia was heavily based on help that was provided by commoner to commoner. At times when a household had no food left, people left their dwellings to travel from house to house asking for food and a place to stay for the night. Thus, gaining help required mobility, which escalated the spread of infectious diseases. The differing degree of socioeconomic and demographic consequences following the 1695 event implies that there is not only one relevant measure of “volcanic impact on society”, but the intensity of the impact depends on the aspects of human life that are investigated.

When comparing the socioeconomic consequences between freeholders and Crown peasants, the deprivation is substantially more profound among the latter group. As the Crown holdings were commonly previously deserted or newly cleared farmsteads on the rural periphery, geographical factors may partly explain the differences. The location of these farmsteads, for example regarding soil properties and microclimate, might have simply been less suitable for crop cultivation than the locations of the freeholder farms . Furthermore, over the century subsidiary livelihood options, like sea and fresh-water fishing or tar burning, became increasingly important in Ostrobothnia. The additional financial assets from these activities may partly explain why certain areas had minor socioeconomic impacts during the 1690s crisis (Fig. d). However, the rights to practice these subsidiary activities were strictly controlled by the authorities, and new licenses were not commonly issued . Thus, the new Crown peasant settlers may not have had the same entitlement to these subsidiary sources for living as the freeholders, who had held the rights generation after generation.

Nonetheless, perhaps the most important factor explaining the differences in terms of socioeconomic consequences between the freeholders and the Crown peasants is immaterial: the continuity of contacts. The peasant families who had settled their holdings over multiple generations had a social status and respect within the community and beyond, which helped in keeping trade and credit relations with burghers of the towns . The mutual trust resulting from these continuous contacts was crucial for the peasants, for example, for gaining loans from the burghers to purchase shipped grain during the 1690s crisis. As an essential element of sustaining these contacts was keeping the holding within one family over generations , newly settled Crown peasant farms or holdings with earlier tax debts were greatly excluded from these arrangements. Continuity may thus have played a role in farm management as well, as examples from other parts of pre-industrial Europe demonstrate . Peasants with hereditary rights may have also been more motivated to invest into the productivity of a farm, and were likely more familiar with local challenges and ways to overcome these than newly settled Crown peasants.

The investigation of the dynamic historical context revealed how different environmental, political, institutional, and cultural factors influenced the socioeconomic consequences of the 17th century eruptions. These components (which are detailed in the previous section) determined the societal and/or individual sensitivity to the adverse cold pulses triggered by volcanic eruptions and the capacity of the local populations to respond to and recover from these events, that is, the degree of vulnerability and resilience (Fig. ).

The concepts of vulnerability and resilience have received critical reassessment by historians over recent years (see, e.g., ). has noted that these two concepts are not opposing or mutually exclusive, but that a resilient society can contain vulnerable people within. The socioeconomic consequences of the 1690s crisis demonstrate this concept: overall, the society seems to be socioeconomically resilient to the 1695 event, but closer investigation reveals one vulnerable societal group, the Crown peasants, among whom the societal hardships accumulated. Likewise, a livelihood system based on climate-sensitive crop cultivation, located in one of the northernmost agricultural areas of Europe, can be considered vulnerable in general. Nevertheless, the people who had access and entitlement to imported grain or could subside their livelihood from alternative resources were more resilient than the others. Thus, vulnerability and resilience coexist within societies, and the influence of these in explaining the societal impacts of volcanic eruptions depends on the scope of our investigation.

The components which influenced the peasant society's vulnerability and resilience were neither spatially nor temporally constant. Furthermore, the effect of these components differed depending on whether, for instance, the demographic or the socioeconomic consequences are investigated. In this regard, whereas the 1601 and the 1641 events had dire socioeconomic consequences, partly due to prevailing political and military circumstances, the tax deferments during the 1690s crisis moderated the ensuing socioeconomic effects. When viewed from a socioeconomic perspective, the authorities' tax relief actions can be seen as a successful institutional coping mechanism. However, when looked at from a demographic perspective, the measures that were directed only to the more advantaged segment of the peasant society had catastrophic consequences. The landless population and the less advantaged Crown peasants that were excluded from the relief measures faced hunger first, and related disease epidemics started to spread, eventually also reaching the wealthier freeholders . Thus, although the authorities' relief actions increased socioeconomic resilience, they also increased the vulnerability to epidemic disease.

The model (Fig. ) was developed from the so-called impact-order model used to detect societal consequences of the 1815 Tambora eruption . The identified components of vulnerability and resilience, however, are relevant to the case study presented in this paper. Likely, alternative components will be identified if similar schemes are created to study the distant socioeconomic consequences of large eruptions within different societies. Nevertheless, it is also likely that some of these components are the same regardless of time and space. For example, similar to Ostrobothnia, in Scotland, poor soil quality, lack of poor-relief institutions, and socio-political distress made the society more vulnerable to the 1690s crisis . Consequently, further micro-regional research focusing on different eruption periods among different societies is needed, if the model is to be improved towards a more universally applicable one.